TTechclick ⚡ XP 0% All lessons
IBM · SIEM · ArchitectureInteractive · L1 / L2 / L3

IBM QRadar Architecture — Console, Processors, QFlow & DSMs

QRadar is not one box — it is a pipeline. Logs and network flows pour in, get parsed and normalized, get correlated into offenses, and land in searchable storage. This lesson maps every component (Console, Event Collector & Processor, QFlow & Flow Processor, the Magistrate, Data Nodes and DSMs) so you can draw the QRadar pipeline and explain events vs flows in an interview.

📅 2026-06-19 · ⏱ 16 min · 5 infographics · live pipeline demo · 🏷 10-Q assessment + AI Tutor inline

⚡ Quick Answer

A clear, interactive guide to IBM QRadar SIEM architecture (2026): all-in-one vs distributed, the Console, the Event Collector and Event Processor (logs), QFlow / Flow Collector and Flow Processor (network flows), the Magistrate that correlates offenses, Data Nodes for storage, how events differ from flows, DSMs that normalize logs, and high-level EPS/FPM licensing.

🎯 By the end you will be able to

Read as:

Pick where you want to start

1

What it is

A pipeline, not a box: all-in-one vs distributed.

2

The event path

Collector, DSMs, Processor, Magistrate, offenses.

3

The flow path

QFlow, Flow Collector & Processor; events vs flows.

4

Storage & sizing

Data Nodes, Ariel store, EPS/FPM licensing.

🧠 Warm-up — 3 questions, no score

Just notice which ones make you pause. We answer all three inside the lesson.

1. Is QRadar a single appliance you drop on the network?

Answered in What it is.

2. What turns a raw firewall log into fields QRadar understands?

Answered in The event path.

3. What does QFlow collect?

Answered in The flow path.

Most engineers think…

Most people picture QRadar as 'a SIEM box where logs go in and alerts come out'. That mental model falls apart the moment an interviewer asks you to draw it or scale it.

QRadar is a distributed pipeline. Logs are gathered by an Event Collector, parsed and normalized by DSMs, then run through rules and stored by an Event Processor. Network traffic is turned into flows by QFlow / Flow Collectors and handled by Flow Processors. The Magistrate on the Console correlates rule hits into offenses, and Data Nodes add storage. You can collapse all of that onto one all-in-one appliance, or spread it across many. Knowing which component does what is exactly what lets you size, troubleshoot and explain QRadar.

① What QRadar actually is — a pipeline, not a box

The single most important idea: QRadar is a pipeline that collects, parses, correlates and stores security data — it is not one inline device. Data enters as two kinds: events (logs from firewalls, servers, AD, cloud) and flows (records of network traffic). Both are normalized into a common shape, run through rules, and saved in a searchable store so analysts can investigate.

QRadar ships in two shapes. An all-in-one deployment runs every role on one appliance. A distributed deployment splits the roles across many appliances — separate Collectors, Processors, Data Nodes and a dedicated Console — so you can scale and place collection close to the data. The current classic line is QRadar SIEM 7.5, sitting alongside IBM's newer cloud-native QRadar SIEM and the broader QRadar Suite (SIEM + SOAR + EDR).

Figure 1 — The QRadar pipeline — collect, parse, correlate, store
Every log and every flow runs the same four-step pipeline before an analyst ever sees it.The QRadar pipeline — collect, parse, correlate, storeCollectEvent & FlowCollectorsParseDSMs normalize fieldsCorrelaterules + MagistrateStoreAriel + Data NodesInvestigateoffenses + search
Every log and every flow runs the same four-step pipeline before an analyst ever sees it.
Figure 2 — All-in-one vs distributed roles
The same roles exist either way — all-in-one stacks them on one box; distributed spreads them out.All-in-one vs distributed rolesConsole + MagistrateUI, offenses, admin and correlationProcessorsEvent & Flow Processors run rules and storeCollectorsEvent Collectors and QFlow gather the data
The same roles exist either way — all-in-one stacks them on one box; distributed spreads them out.
Quick check · Q1 of 10 · Understand

Which best describes QRadar's architecture?

Correct: b. QRadar is a pipeline of roles — Collectors, Processors, Magistrate, Data Nodes — that you can stack on one all-in-one box or spread across many appliances. It is not a single inline blocking device.
👉 So far: QRadar is a collect-parse-correlate-store pipeline — runnable all-in-one (one box) or distributed (many appliances), with the same roles either way.

② The event path — Collector, DSMs, Processor and the Magistrate

Follow a log end to end. An Event Collector receives raw logs (syslog, WinCollect, APIs), queues them, and hands them to a DSM that parses and normalizes the message into common fields: Event ID, Source IP, Destination IP, Username, Protocol and more. It also coalesces near-identical events to cut volume. The Event Processor then runs the normalized events through the Custom Rules Engine (CRE) and writes them to storage.

Where offenses come from

When rules fire, the Processors send notifications up to the Magistrate (MPC), which lives only on the Console or all-in-one. The Magistrate correlates those hits across all Processors and raises a single offense — the prioritized case an analyst works. In distributed mode the Console does no event/flow processing itself; it is the UI plus the Magistrate.

Figure 3 — The Magistrate correlates from every Processor
Each Processor reports rule hits to the single Magistrate on the Console, which raises offenses.The Magistrate correlates from every ProcessorMagistrateon the ConsoleEvent Processor AEvent Processor BFlow ProcessorEvent CollectorQFlow CollectorData Node
Each Processor reports rule hits to the single Magistrate on the Console, which raises offenses.
Figure 4 — From raw log to offense
A single log is parsed, normalized, run through rules and correlated into a prioritized offense.From raw log to offenseRaw logsyslog / API inDSM parsenormalize fieldsCRE rulesEvent ProcessorMagistratecorrelate hitsOffenseanalyst case
A single log is parsed, normalized, run through rules and correlated into a prioritized offense.
🖥️
Console
tap to flip

The web UI for searches, reports and offenses — and the only place the Magistrate runs. In distributed mode it does no event or flow processing itself.

📥
Event Collector + DSM
tap to flip

Gathers raw logs and uses a DSM to parse and normalize them into common fields (Source IP, Username, Event ID), then coalesces and forwards.

🌐
QFlow / Flow Processor
tap to flip

QFlow turns network traffic into flow records (bytes, packets, ports, duration); the Flow Processor runs flow rules and stores them.

🗄️
Data Node
tap to flip

Bolts onto a Processor to add storage and search capacity for the Ariel store, so you scale retention without swapping appliances.

Name the role, not just 'QRadar'

In an interview, separate collection (Event Collector + DSMs, QFlow) from processing (Event/Flow Processors run rules and store) from correlation (the Magistrate on the Console raises offenses). Saying 'the Event Processor stores and the Magistrate correlates' instantly signals you understand the pipeline.

Quick check · Q2 of 10 · Remember

What component parses a raw vendor log into normalized QRadar fields?

Correct: c. A DSM knows a specific log format and maps its fields to QRadar properties like Source IP, Username and Event ID. The Magistrate correlates offenses; Data Nodes store; the Flow Processor handles flows.
👉 So far: Event path: Collector gathers logs → DSM parses and normalizes → Event Processor runs CRE rules and stores → Magistrate on the Console correlates hits into offenses.

③ The flow path — QFlow, Flow Processor, and events vs flows

Network traffic takes a parallel path. QFlow / Flow Collectors read traffic (from a SPAN/mirror or tap, or by ingesting NetFlow/IPFIX/sFlow) and build flow records — who talked to whom, on which ports, for how long, and how many bytes and packets. A Flow Processor processes those flows, runs flow rules, and stores them, the same way the Event Processor handles logs.

The interview line — events vs flows: an event is a point-in-time log of one action (a login, a VPN connect, a deny). A flow is a record of a network session that can last seconds to hours, summarising bytes, packets and ports between two hosts. Events tell you what a device reported; flows tell you what actually moved on the wire — including traffic no device logged.

Figure 5 — Events vs flows
Two data types, two paths — events are logged actions; flows are network sessions on the wire.Events vs flowsEvents (logs)Point-in-time actionFrom firewalls, AD, servers, cloudParsed by DSMsLicensed in EPSSays what a device reportedFlows (network)A session over timeBuilt by QFlow from trafficBytes, packets, ports, durationLicensed in FPMSays what moved on the wire
Two data types, two paths — events are logged actions; flows are network sessions on the wire.
'Events and flows are the same' confusion

They are not. An event is a logged action a device reported; a flow is a record of an actual network session (bytes, packets, ports) built by QFlow — including traffic no device ever logged. Always answer with both paths and license each separately (EPS vs FPM).

▶ Watch a firewall login log become an offense

How one raw log travels the QRadar pipeline end-to-end. Press Play for the healthy path, then Break it to see the classic failure.

① CollectA firewall sends a syslog 'admin login from a new country' message to the Event Collector.
② ParseThe matching DSM parses and normalizes it — Source IP, Username and Event ID are now real QRadar fields.
③ CorrelateThe Event Processor's CRE matches a rule; the hit goes up to the Magistrate, which correlates it with related activity.
④ OffenseThe Magistrate raises a single offense; the analyst opens it on the Console with full context.
Press Play to step through the healthy pipeline. Then press Break it.
Quick check · Q3 of 10 · Understand

What is the key difference between an event and a flow?

Correct: a. An event is one logged action at a moment (a login, a deny). A flow summarises a network session — bytes, packets, ports, duration — and can span seconds to hours. Events are what devices reported; flows are what moved on the wire.
👉 So far: Flow path: QFlow / Flow Collector builds flow records (bytes, packets, ports, duration) → Flow Processor processes them. Events = logged actions; flows = network sessions.

④ Storage and sizing — Data Nodes, the Ariel store and EPS/FPM

Processors write events and flows to the Ariel time-series store (offenses live in a Postgres database). When you need more retention or faster search, you add Data Nodes: they attach to Processors, take over a share of storage and search, and let you scale capacity without replacing appliances. Searches run over Ariel using AQL (covered in the next lesson).

How it is licensed

Capacity is licensed by throughput: EPS (events per second) for logs and FPM (flows per minute) for flows. Licenses apply to Processors, not Collectors — data is collected anywhere but counted where it is processed. In a distributed estate you pool capacity and distribute it to Processors, adding Processors and Data Nodes as volume grows. The classic mistake is sizing only for today's EPS and ignoring flow volume and retention.

Vikram, a SOC engineer at a Pune fintech, faces this

A new firewall is onboarded, but its logs show up in QRadar as 'Unknown' events with empty Source IP and Username fields, so no rules fire on them.

Likely cause

QRadar has no matching DSM for that firewall's log format, so the events arrive raw and unparsed — nothing is normalized.

Diagnosis

Open the Log Activity tab — the events are categorised as Unknown / Stored, not parsed; the log source has no DSM mapping its fields.

Admin ▸ DSM Editor + Log Source Management ▸ Log Sources
Fix

Install or auto-update the correct DSM (or build a custom parser in the DSM Editor) so the fields normalize, then confirm the log source type is set correctly.

Verify

Re-check Log Activity: the same logs now show a proper event name, Source IP and Username — and the rules that target them start firing.

Prove parsing from Log Activity, not a hunch

Never assume a log source 'is probably fine'. The Log Activity tab shows whether events are normalized or sitting as Unknown. One look tells you if the DSM is doing its job — no guessing about why a rule did not fire.

Quick check · Q4 of 10 · Apply

Your log volume is climbing and searches are slow. What is the right scaling move?

Correct: b. You scale QRadar horizontally — add Data Nodes for storage/search and Processors for throughput, and license more EPS/FPM. Shrinking visibility by disabling parsers or flows defeats the purpose.
👉 So far: Processors store to the Ariel store; Data Nodes add storage and search. License logs in EPS and flows in FPM, applied to Processors — scale by adding Processors and Data Nodes.

🤖 Ask the AI Tutor

Tap any question — instant, scoped to this lesson. No login, no waiting.

Pre-curated from vendor docs + community Q&A, scoped to this lesson. For a live prod issue, paste your export into chat.techclick.in.

📝 Wrap-up assessment — six more

You've answered 4 inline. Six left. 70% (7 of 10) marks the lesson complete on your profile. Tap Submit all answers at the end.

Q5 · Remember

Which component runs the Magistrate and correlates offenses?

Correct: b. The Magistrate (MPC) runs only on the Console or an all-in-one appliance. It correlates rule hits from all Processors into offenses. Collectors gather data; Data Nodes store; QFlow builds flows.
Q6 · Understand

In a distributed deployment, what does the Console stop doing?

Correct: a. In distributed mode the Console is the UI plus the Magistrate; it does not perform event/flow processing or storage — that is handled by the dedicated Processors and Data Nodes.
Q7 · Apply

A new log source shows up as 'Unknown' events with empty fields. What is the fix?

Correct: a. 'Unknown' events mean no DSM matched, so nothing was parsed. Installing or building the right DSM (in the DSM Editor) normalizes the fields so rules can fire. Storage and flow licensing are unrelated to parsing.
Q8 · Analyze

Why might QRadar see malicious traffic that no log source ever recorded?

Correct: c. Flows are built by QFlow from the traffic itself (bytes, packets, ports), so they reveal sessions even when no device produced a log. That is the core value of having both events and flows.
Q9 · Evaluate

An interviewer asks how to scale QRadar for far more log volume and longer retention. Best answer?

Correct: d. You scale horizontally: add Processors for throughput and Data Nodes for storage/search, distribute licensed EPS/FPM across Processors, and keep one Console with the Magistrate. A single bigger box and disabling visibility are the wrong calls.
Q10 · Evaluate

What is the correct way to think about EPS and FPM licensing?

Correct: d. EPS (events per second) covers logs and FPM (flows per minute) covers flows. Both are applied to the Processors that process and store the data, not to the Collectors that merely gather it. Flows are licensed, not free.
Lesson complete — saved to your profile.
Almost! You need 70% (7 of 10) — re-read the path that tripped you up and tap "Try again".

🧠 In your own words

Type one line: why is QRadar called a pipeline rather than a box? Then compare with the expert version.

Expert version: Because data flows through distinct roles in order: Event Collectors and QFlow gather logs and flows, DSMs parse and normalize them, Event and Flow Processors run rules and store the data, the Magistrate on the Console correlates rule hits into offenses, and Data Nodes add storage and search. You can stack every role on one all-in-one appliance or spread them across many in a distributed deployment, and you scale by adding Processors and Data Nodes and licensing more EPS and FPM. There is no single inline 'QRadar box' — there is a pipeline of components, which is exactly why you can size and troubleshoot each stage independently.

🗣 Teach a friend

Best way to lock it in — explain it in one line to a teammate. Tap to generate a paste-ready summary.

📖 Glossary

Console
The QRadar web UI for searches, reports and offenses, and the only place the Magistrate runs. In distributed mode it does no event/flow processing.
Event Collector
Gathers raw logs from sources (syslog, WinCollect, APIs), queues them, applies DSM parsing/normalization and coalescing, then forwards to a Processor.
Event Processor
Runs normalized events through the Custom Rules Engine (CRE), executes rule responses, and stores events in the Ariel store.
QFlow / Flow Collector
Builds network flow records from traffic (SPAN/tap) or ingested NetFlow/IPFIX/sFlow — bytes, packets, ports and duration per session.
Flow Processor
Processes flow records, runs flow rules and stores flows — the flow-side counterpart of the Event Processor.
Magistrate (MPC)
The Magistrate Processing Core on the Console; correlates rule hits from all Processors and raises prioritized offenses.
Data Node
An appliance that attaches to a Processor to add storage and search capacity for the Ariel store, scaling retention horizontally.
DSM (Device Support Module)
A parser for a specific log source format that normalizes raw logs into common QRadar fields like Source IP, Username and Event ID.
Event vs flow
An event is a point-in-time logged action; a flow is a record of a network session over time (bytes, packets, ports, duration).
EPS / FPM
Licensing throughput metrics — events per second for logs and flows per minute for flows — applied to Processors.

📚 Sources

  1. IBM Documentation — QRadar events and flows (overview). ibm.com/docs/en/qsip/7.5.0 ▸ overview-qradar-events-flows
  2. IBM Documentation — QRadar 7.5 Architecture and Deployment Guide (Console, Event/Flow Processors, Magistrate, Data Nodes). ibm.com/docs/en/SS42VS_7.5
  3. IBM Documentation — Distributing event and flow capacity; EPS and FPM licensing. ibm.com/docs/en/qsip/7.5.0 ▸ management-distributing-event-flow-capacity
  4. IBM Support — QRadar: About EPS & FPM Limits. ibm.com/support/pages/qradar-about-eps-fpm-limits
  5. IBM Documentation (DSM) — Introduction to log source management and supported DSMs. ibm.com/docs/en/dsm
  6. IBM — Cloud-native QRadar SIEM and the QRadar Suite (SIEM + SOAR + EDR). ibm.com/products/qradar-siem

What's next?

Got the architecture? Next, go deep on what QRadar does with all that data: rules and building blocks in the Custom Rules Engine, how offenses get a magnitude, AQL searches over the Ariel store, and User Behavior Analytics (UBA).